Use of operative gin config #968
Unanswered
marcospiau
asked this question in
General
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi,
In case I want to continue pre-training using another dataset/vocab is it necessary to pass the operative config used on pre-training or simply pointing to the correct checkpoint path is enough? I'm training via command line using the t5_mesh_transformer command.
Thanks,
Marcos
Beta Was this translation helpful? Give feedback.
All reactions